USE OF NEURAL NETWORKS FOR ANNOTATING SEARCH RESULTS

• US 20110047111A1
• Filed: 11/15/2009
• Published: 02/24/2011
• Est. Priority Date: 09/26/2005
• Status: Active Grant

• Current Assignee: Callahan Cellular LLC (Intellectual Ventures LLC)
• Original Assignee: Quintura Inc.
• Inventors: Ershov, Alexander V.
• Granted Patent: US 8,533,130 B2
• Time in Patent Office: Days
• US Class Current: 706/20
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

A system for generating annotations of a document, including a plurality of neurons connected as a neural network, the neurons being associated with words, sentences and documents. An activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. The neurons are displayed to a user and identify the neurons that correspond to sentences containing a predetermined percentage of document meaning. The annotations can be also based on a context of the user'"'"'s search query. The query can include keywords, documents considered relevant by the user, or both. Positions of the neurons relative to each other can be changed on a display device, based on input from the user, with the change in position of one neuron changing the resulting annotations. The input from the user can also include changing a relevance of neurons relative to each other, or indicating relevance or irrelevance of a document or sentence.

Claims

1. A system for generating annotations of a document, the system comprising:

a) a processor;

b) a memory coupled to the processor;

c) computer code loaded into the memory for implementing the following functionality;

d) a plurality of neurons organized as a neural network, the neurons being associated with words (word-neurons) and sentences (sentence-neurons), the sentences being extracted from documents,

e) wherein the word-neurons are organized into an input layer that receives an input query, and the sentence-neurons are organized into a sentence layer,

f) wherein at least some of the word-neurons of the input layer have connections between each other; and

g) in response to a user interactively changing a context of the input query, means for displaying sentences corresponding to the sentence-neurons to a user and identifying the sentence-neurons that correspond to sentences taken from the documents having the highest relevance to document meaning based on a predetermined percentage of document meaning,

h) wherein the predetermined percentage of document meaning is a function of a percentage of meaning of the displayed sentences;

i) wherein the neural network calculates the percentage of meaning of each sentence, describing the degree of relevance of the sentence to the document, and

j) wherein the taken sentences represent annotations of a selected document.

2. The system of claim 1, wherein the annotations are also based on a context of the user'"'"'s search query.

3. The system of claim 1, wherein the neural network is a bidirectional neural network.

4. The system of claim 1, wherein the input query comprises documents considered relevant by the user.

5. The system of claim 1, further comprising means for changing positions of the word-neurons on a display relative to each other based on input from the user, wherein the change in position of one word-neuron changes the annotations.

6. The system of claim 1, wherein the input from the user comprises dragging a word on a display device.

7. The system of claim 1, wherein the input from the user comprises changing a relevance of neurons relative to each other.

8. The system of claim 1, wherein the user can inhibit neurons of the neural network by indicating irrelevance of a sentence neuron.

9. The system of claim 1, wherein the neural network is excited by a query that identifies a plurality of documents considered relevant by a user.

10. The system of claim 1, further comprising an activity regulator that regulates a sum of all activity of all active neurons of the neural network that are excited at any given time, and wherein the activity regulator regulates at least one of a minimum activity level of the neural network and a maximum activity level of the neural network.

11. The system of claim 1, wherein the word-neurons include additional words, different from the input query, generated by the neural network from the sentences.

12. The system of claim 1, wherein the neural network is bidirectional.

13. A system for generating annotations of a document, the system comprising:

a) a processor;

b) a memory coupled to the processor;

c) computer code loaded into the memory for implementing the following functionality;

d) a plurality of neurons organized as a neural network, the neurons being associated with words (word-neurons) and sentences (sentence-neurons), the sentences being extracted from documents,

e) wherein the word-neurons are organized into an input layer that receives an input query, and the sentence-neurons are organized into a sentence layer,

f) wherein the word-neurons include additional words, different from the input query, generated by the neural network from the sentences; and

g) in response to a user interactively changing a context of the input query, means for displaying sentences corresponding to the sentence-neurons to a user;

h) wherein the neural network calculates the percentage of meaning of each sentence, describing the degree of relevance of the sentence to the document, and

i) wherein the taken sentences represent annotations of a selected document; and

j) an activity regulator that regulates a sum of all activity of all active neurons of the neural network that are excited at any given time, and wherein the activity regulator regulates at least one of a minimum activity level of the neural network and a maximum activity level of the neural network.

14. The system of claim 13, wherein at least some of the word-neurons of the input layer have connections between each other.

15. The system of claim 13, further comprising means for identifying the sentence-neurons that correspond to sentences taken from the documents having the highest relevance to document meaning based on a predetermined percentage of document meaning, wherein the predetermined percentage of document meaning is a function of a percentage of meaning of the displayed sentences

16. A computer implementable method for generating annotations, the method being performed on a computer having a processor and a memory, the method for performing the steps of:

(a) forming a neural network from a plurality of neurons, the neurons being associated with words (word-neurons) and sentences (sentence-neurons), the sentences being extracted from documents,

(b) organizing the word-neurons into an input layer that receives an input query, and organizing the sentence-neurons into a sentence layer,

(c) wherein at least some of the word-neurons of the input layer have connections between each other; and

(d) in response to a user interactively changing a context of the input query, displaying sentences corresponding to the sentence-neurons to a user and identifying the sentence-neurons that correspond to sentences taken from the documents having the highest relevance to document meaning based on a predetermined percentage of document meaning,

(e) wherein the predetermined percentage of document meaning is a function of a percentage of meaning of the displayed sentences;

(f) using the neural network, calculating the percentage of meaning of each sentence, describing the degree of relevance of the sentence to the document, and

(g) wherein the taken sentences represent annotations of a selected document.

17. The method of claim 16, wherein the annotations are also based on a context of the input query.

18. The method of claim 16, further comprising means for changing positions of the word-neurons on a display relative to each other based on input from the user, wherein the change in position of one word-neuron changes the annotations.

19. The method of claim 16, further comprising regulating a sum of all activity of all active neurons of the neural network that are excited at any given time, and including regulating at least one of a minimum activity level of the neural network and a maximum activity level of the neural network.

20. A computer implementable method for generating annotations, the method being performed on a computer having a processor and a memory, the method for performing the steps of:

(a) organizing a plurality of neurons into a neural network, the neurons being associated with words (word-neurons) and sentences (sentence-neurons), the sentences being extracted from documents,

(b) organizing the word-neurons into an input layer that receives an input query, and organizing the sentence-neurons are organized into a sentence layer,

(c) wherein the word-neurons include additional words, different from the input query, generated by the neural network from the sentences;

(d) in response to a user interactively changing a context of the input query, displaying sentences corresponding to the sentence-neurons to a user;

(e) using the neural network, calculating the percentage of meaning of each sentence, describing the degree of relevance of the sentence to the document, and

(f) wherein the taken sentences represent annotations of a selected document; and

(g) regulating a sum of all activity of all active neurons of the neural network that are excited at any given time, and including regulating at least one of a minimum activity level of the neural network and a maximum activity level of the neural network.